Hugging Face launched Carbon-X, a free AI model that reads genetic code and answers questions about DNA in plain language.
Image source: ForkLog
Hugging Face, a company that builds artificial intelligence (AI) tools, has released a new model called Carbon-X that can read human and animal DNA and answer questions about it in plain English — no programming skills required.
DNA is the "instruction manual" inside living things. It's written using just four letters: A, C, G, and T. Carbon-X was trained to read these letters in groups of six, so it can understand genetic code the same way it understands normal text.
What can it do? The model helps scientists with medical and biological research. For example, it can:
The model is powerful but needs serious hardware. It has 14.8 billion "parameters" (the internal settings an AI uses to make decisions) and can run locally — meaning sensitive genetic data stays on the user's own computer instead of being sent to outside servers. But running it requires a high-end graphics processor with around 80 GB of memory.
In tests, Carbon-X performed well against expensive rivals:
Important limits: Carbon-X has not passed clinical testing, so its predictions must not be used for real medical decisions. It can also make mistakes counting DNA units and sometimes "invents" answers when it lacks context.
Carbon-X follows an earlier model, Carbon-A, which helped build a database of 566 million possible genes across more than 22,000 species.
This is an AI-generated summary. Read the original article at: https://forklog.com/news/ai/hugging-face-nauchila-ii-obshhatsya-s-dnk